Research On Dehazing Algorithm Based On Single Image

With the rapid development of the country,air quality problems have become more and more serious,especially the problem of smog.The image acquired in such weather is of poor quality and does not better highlight the details of the image.Therefore,how to better remove the smog information and obtain more image details has become an urgent problem to be solved.Aiming at the above problems,this paper firstly simulates various defogging algorithms based on image enhancement to achieve defogging,and analyzes the defogging effect from subjective evaluation criteria and objective evaluation criteria.The objective evaluation indicators selected in this paper are mainly effective.Detail intensity,degree of tonal reduction,information entropy,peak signal to noise ratio,and run time.The traditional single-scale Retinex algorithm and multi-scale Retinex algorithm are studied,and the problem of easy generation of halo effect and huge computational complexity is improved.Firstly,the image is spatially transformed from RGB space to HIS space.Then the wavelet transform is used to process the luminance component into high and low frequency regions.In the low frequency region,the improved Retinex algorithm is used to process it.The frequency region is processed by high frequency coefficients and sharpening.In order to better improve image contrast,global contrast enhancement will be performed using the improved PCNN model.At the same time,in order to ensure that the color of the image after restoration is more natural,the saturation enhancement will be performed synchronously,and after the spatial inverse transformation,the dynamic range stretching will be performed.Finally,the traditional Retinex algorithm and sub-block overlapping histogram equalization algorithm are compared from the subjective and objective evaluation criteria,which shows the superiority of the improved algorithm.In addition,the whole defogging process of the dark primary color prior algorithm is introduced.Although the algorithm is very classic,there are still some shortcomings in the estimation of atmospheric light value,the optimization of transmittance and the efficiency of defogging.Therefore,these problems are mainly improved.And from the subjective evaluation criteria and the objective evaluation criteria,it is compared with several other methods based on the dark primary color prior image dehazing algorithm.In order to ensure the validity of the conclusion,a large number of foggy images were randomly selected for experiment,and the results were compared with the other two algorithms from the three aspects of effective detail intensity,hue reduction degree and defogging time.The improved algorithm has better defogging effect.